New Phase-Lead Compensator / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

New Phase-Lead Compensator

1 2 next

Move Thread

LAN_403

dude_500

Sun Oct 02 2011, 03:21AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

This morning I woke up after a bad dream regarding phase leading, and I decided I no longer wanted to put up with the first-crossing-error that always comes up with a simple inductive phase-leader in an RL burden network (which I still don't understand why it has an error, if you solve the differential equation it says that the network should phase-lead instantaneously upon any harmonic input oscillation).

So I just emulated the RL network using op-amps... invert the input, take the derivative times a negative constant A, and add that to the original inverted input times constant B. Invert this output, and you get the input scaled with an angular offset of tan-1(wA/B). This behaves much better than the inductor RL network since it seems the op-amp can establish the derivative faster than the inductor will take on L dI/dt after an impulse. I'm getting flawless phase lead on even the first half-cycle crossing, and my op-amps aren't even anything special... using TL072 in this case though I might pick up some with slightly a slightly higher unity gain (though I really don't even need to).

I don't feel like plugging in values (and it still could use some tweaking anyways), so if you want to use this in your designs, calculate your own values (since it'll vary by coil anyways since the offset depends on frequency).

As can be seen in the video, the first cycle phase is accurate and in addition, amplitude has no impact on lead. These are the two constraints of a perfect DRSSTC phase lead compensator.

I've got a bunch of midterms next week, but hopefully I'll get to try it out on my big coil within a week or two if some extra time comes up.

Schematic:

Goodchild

Sun Oct 02 2011, 05:35AM

Registered Member #2292 Joined: Fri Aug 14 2009, 05:33PM
Location: The Wild West AKA Arizona
Posts: 795

My theory to why the phase lead doesn't work on the first cycle is this: Phase lead works by predicting where the voltage will be based on "previous" information. If there is no previous information or a lack of it (like only a half a cycle) then the phase lead can't accurately predict where the waveform will be. So at the start of a pulse where there is no or little information to base your prediction on it has no choice but to flow the input directly until it accumulates and can start predicting.

As in real life you need a "lot" of past information to make a good prediction. Yes it will start predicting right as it starts but it only has a small part of the wave to work with so it's not going to be very accurate.

Sorry for the rant just thought I would put that out for discussion.

Just wondering but why go to so much trouble to get rid of that first spike? It usually dissipates with higher power anyway and doesn't cause any need for concern in any of my systems. Also an all pass filter may have a similar effect and use less parts.

Great work though! I look foreword to seeing if it dose get rid of that first spike. I really love seeing thought provoking post in the forum, usually it's just stuff like "HELP MY SGTC DON'T WORK" hehe

Eric

ScotchTapeLord

Sun Oct 02 2011, 07:39AM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

Good thinking, but I find a lot of your inversions redundant.

I have simplified your circuit. Here is the simulation:

I look forward to testing it for myself.

Please note that the first one is an op amp, and the second is a fast comparator (this feeds the DRSSTC logic circuit).
I would also include, along with clamping diodes, a resistor in series with the capacitor to reduce high frequency noise amplification, which may be an issue with both of our circuits. I would even consider a low pass filter.
The component values are not optimized in mine, either, but aren't terribly critical.

Hope I helped!

dude_500

Sun Oct 02 2011, 08:10AM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

ScotchTapeLord wrote ...

Good thinking, but I find a lot of your inversions redundant.

I have simplified your circuit. Here is the simulation:

The first op-amp in my schematic is necessary because as far as my experimentation shows, a differentiator can't be fed from a high-impedance source. I'm not sure why you're dividing your source through a 2000ohm - 5ohm divider, but I believe that is creating a low impedance connection to your source since you're so close to ground, but in reality if you've run it through a resistor in order to safely clamp, it will be high impedance.

axelro

Sun Oct 02 2011, 10:32AM

Registered Member #3640 Joined: Sat Jan 22 2011, 12:16PM
Location: Germany close to Heidelberg
Posts: 39

Argg - after playing around quite some time, I can't follow these results in LTSpice, neither yours, Dude_500, nor yours, ScotchTapeLord (in your case I rather see phase lag). Unfortunately I lack the knowledge to calculate this on my own.

Dude_500, I know you stated you feel like plugging in values, but would you eventually be willing to at least share the values you used in your test circuit as a starting point, please?

I did actually try all pass circuits (2nd order), but this is not giving a good 1st cycle signal at all, and it's introducing a huge propagation delay.I am currently using a Piccolo uC from TI to try introduce the phase lead digitally.

Anyway, attaching the LTSpice simulations, as a minor contribution.

]phase_shift_dude_500t.zip[/file]
]phase_shift_scotchtapelord.zip[/file]

ScotchTapeLord

Sun Oct 02 2011, 04:04PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

I was approximating a current source with a high voltage source feeding a high resistance source, but the highest voltage I could use was 1kV so I was limited.

I would think as long as the differentiator's input impedence (determined by the R and C) is much larger than your burden resistor that it should work, but if your work shows otherwise, just include a unity gain buffer after the burden resistor. It won't increase the component count at all because it would just require a dual op amp.

dude_500

Sun Oct 02 2011, 04:22PM

Registered Member #2288 Joined: Wed Aug 12 2009, 10:42PM
Location: Cambridge, MA
Posts: 179

axelro wrote ...

Dude_500, I know you stated you feel like plugging in values, but would you eventually be willing to at least share the values you used in your test circuit as a starting point, please?

Here is an ltspice simulation that works. Although, in actual construction some of the part values are considerably different (by orders of magnitude in some cases). These are just good values for the op-amps in ltspice and the simulation system.
]phase_lead.zip[/file]

ScotchTapeLord

Sun Oct 02 2011, 05:00PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

axelro wrote ...

Argg - after playing around quite some time, I can't follow these results in LTSpice, neither yours, Dude_500, nor yours, ScotchTapeLord (in your case I rather see phase lag). Unfortunately I lack the knowledge to calculate this on my own.

Your midpoint tap needs a large value resistor (50k or so) going to the node where R4 and C1 meet. Disconnect the CT and burden from the midpoint tap and ground them. Also it's important to make sure your op amp can run at these voltages. You need to ground the bottom of your CT, too. The values might not work for your supply voltage, as my simulations used +/-15V supplies.

My design needs refinement, for sure, the more I look at it. Late night circuit design...

EDIT: Here is a much better realization of the circuit than my last, staying much truer to the original design by dude,

The differentiator should have its R and C set for a gain of 1 for your coil's frequency. Also, its negative feedback resistor needs to be significantly larger than the input resistor before the 3nF capacitor for maximum phase shift capability (which won't necessarily be necessary). The potentiometer should probably be pretty high, maybe 100k. It isn't dramatically different from the original design at all, but I needed to redeem myself after that last trainwreck that I posted.

Finn Hammer

Sun Oct 02 2011, 08:14PM

Registered Member #205 Joined: Sat Feb 18 2006, 11:59AM
Location: SkÃ¸rping, Denmark
Posts: 741

All,

I have not had time to read all replies in detail, but it is work like yours that make me feel inclined to join into the loop again and upgrade the Predikter. The predictor should be elegant, and it is only totally so if it provides accurate phase lead on e-ve-ry crossing, even the first.

So: Great work, there is hope still.

Cheers, Finn Hammer

ScotchTapeLord

Sun Oct 02 2011, 09:21PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

While I think the improvements are good from a design perspective, the slight error caused during the first cycle is probably negligible in the face of the losses it eliminates in the later cycles when there is much more current flowing.

I think these new circuits are good because they use only a few components and there is only one knob to turn, and there is no inductor to wind and shield.

Thinking about my design, I'd definitely increase the RC feedback network values so that gain varies less with frequency.

1 2 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob